+
+> **⚡ANSR Training on Fully Procedurally Generated Data** Inspired by NeSymReS ([Biggio et al. 2021](https://arxiv.org/abs/2106.06427))
+
# Introduction
+TODO
+
+
+
+> **Model Comparison.** Up to 3 variables. Default Model Configurations (32 threads / beams).\
+> Bootstrapped Median, 5p, 95p and AR-p ([Noreen 1989](https://scholar.google.com/scholar?hl=en&q=Computer-intensive+methods+for+testing+hypotheses)) values (n=1000).\
+> N = 5000 ([⚡ v7.0](#usage)), 1000 ([PySR](https://github.com/MilesCranmer/PySR), [NeSymReS 100M](https://github.com/SymposiumOrganization/NeuralSymbolicRegressionThatScales?tab=readme-ov-file#pretrained-models)).\
+> AMD 9950X (16C32T), RTX 4090.
+
+# Table of Contents
+- [Introduction](#introduction)
+- [Table of Contents](#table-of-contents)
+- [Requirements](#requirements)
+ - [Hardware](#hardware)
+ - [Software](#software)
+- [Getting Started](#getting-started)
+ - [1. Clone the repository](#1-clone-the-repository)
+ - [2. Install the package](#2-install-the-package)
+- [Usage](#usage)
+- [Training](#training)
+ - [Express](#express)
+ - [Manual](#manual)
+ - [0. Prerequisites](#0-prerequisites)
+ - [1. Import test data](#1-import-test-data)
+ - [2. Generate validation data](#2-generate-validation-data)
+ - [3. Train the model](#3-train-the-model)
+ - [4. Evaluate the model](#4-evaluate-the-model)
+ - [4.1 Evaluate NeSymReS](#41-evaluate-nesymres)
+ - [4.2 Evaluate PySR](#42-evaluate-pysr)
+- [Development](#development)
+ - [Setup](#setup)
+ - [Tests](#tests)
+- [Citation](#citation)
+
# Requirements
## Hardware
@@ -37,7 +73,7 @@ cd flash-ansr
## 2. Install the package
-Optional: Create a virtual environment:
+Create a virtual environment (optional):
**conda:**
@@ -50,13 +86,11 @@ Then, install the package via
```sh
pip install -e .
-pip install -e nsrops
+pip install -e ./nsrops
```
# Usage
-## Use a pre-trained model
-
```python
import torch
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@@ -227,12 +261,12 @@ with
- `-o` the output file for results
- `-v` verbose output
-#### 4.1 Evaluate NeSymRes
+#### 4.1 Evaluate NeSymReS
1. Clone [NeuralSymbolicRegressionThatScales](https://github.com/SymposiumOrganization/NeuralSymbolicRegressionThatScales) to a directory of your choice.
2. Download the `100M` model as described [here](https://github.com/SymposiumOrganization/NeuralSymbolicRegressionThatScales?tab=readme-ov-file#pretrained-models)
3. Move the `100M` model into `flash-ansr/models/nesymres/`
4. Create a Python 3.10 (!) environment and install flash-ansr as in the previous steps.
-5. Install NeSymRes in the same environment:
+5. Install NeSymReS in the same environment:
```sh
cd NeuralSymbolicRegressionThatScales
pip install -e src/
@@ -264,13 +298,7 @@ pre-commit install
## Tests
-Test the package with
-
-```sh
-./scripts/pytest.sh
-```
-
-for convenience.
+Test the package with `./scripts/pytest.sh`. Run pylint with `./scripts/pylint.sh`.
# Citation
```bibtex
@@ -279,7 +307,7 @@ for convenience.
title = {Flash Amortized Neural Symbolic Regression},
year = 2024,
publisher = {GitHub},
- version = {0.1.0},
+ version = {0.3.0},
url = {https://github.com/psaegert/flash-ansr}
}
```
diff --git a/assets/images/nsr-training.drawio.svg b/assets/images/nsr-training.drawio.svg
new file mode 100644
index 0000000..e17c96d
--- /dev/null
+++ b/assets/images/nsr-training.drawio.svg
@@ -0,0 +1,4 @@
+
+
+
+
\ No newline at end of file
diff --git a/assets/images/results.png b/assets/images/results.png
new file mode 100644
index 0000000..cc4eb5e
Binary files /dev/null and b/assets/images/results.png differ
diff --git a/configs/v10.0/dataset_train.yaml b/configs/v10.0/dataset_train.yaml
new file mode 100644
index 0000000..bb27322
--- /dev/null
+++ b/configs/v10.0/dataset_train.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_train.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v10.0/dataset_val.yaml b/configs/v10.0/dataset_val.yaml
new file mode 100644
index 0000000..c15ed03
--- /dev/null
+++ b/configs/v10.0/dataset_val.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_val.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v10.0/evaluation.yaml b/configs/v10.0/evaluation.yaml
new file mode 100644
index 0000000..98f6576
--- /dev/null
+++ b/configs/v10.0/evaluation.yaml
@@ -0,0 +1,14 @@
+n_support: 512
+beam_width: 32
+n_restarts: 8
+max_len: 32
+refiner_method: 'curve_fit_lm'
+numeric_head: False
+equivalence_pruning: True
+pointwise_close_criterion: 0.95
+pointwise_close_accuracy_rtol: 0.05
+pointwise_close_accuracy_atol: 0.001
+r2_close_criterion: 0.95
+refiner_p0_noise: 'uniform'
+refiner_p0_noise_kwargs: {'low': -5, 'high': 5}
+device: cuda
\ No newline at end of file
diff --git a/configs/v10.0/expression_space.yaml b/configs/v10.0/expression_space.yaml
new file mode 100644
index 0000000..8b17a96
--- /dev/null
+++ b/configs/v10.0/expression_space.yaml
@@ -0,0 +1,256 @@
+operators:
+ "+":
+ realization: "+"
+ alias: ["add", "plus"]
+ inverse: "-"
+ arity: 2
+ weight: 10
+ precedence: 1
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "-":
+ realization: "-"
+ alias: ["sub", "minus"]
+ inverse: "+"
+ arity: 2
+ weight: 5
+ precedence: 1
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ neg:
+ realization: "nsrops.neg"
+ alias: ["negative"]
+ inverse: "neg"
+ arity: 1
+ weight: 5
+ precedence: 2.5
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ "*":
+ realization: "*"
+ alias: ["mul", "times"]
+ inverse: "/"
+ arity: 2
+ weight: 10
+ precedence: 2
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "/":
+ realization: "/"
+ alias: ["div", "divide"]
+ inverse: "*"
+ arity: 2
+ weight: 5
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ abs:
+ realization: "abs"
+ alias: ["absolute"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ inv:
+ realization: "nsrops.inv"
+ alias: ["inverse"]
+ inverse: "inv"
+ arity: 1
+ weight: 4
+ precedence: 4
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ pow2:
+ realization: "nsrops.pow2"
+ alias: ["square"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ pow3:
+ realization: "nsrops.pow3"
+ alias: ["cube"]
+ inverse: "pow1_3"
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow4:
+ realization: "nsrops.pow4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 1
+ positive: true
+ monotonicity: 0
+ pow5:
+ realization: "nsrops.pow5"
+ alias: []
+ inverse: "pow1_5"
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_2:
+ realization: "nsrops.pow1_2"
+ alias: ["sqrt"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_3:
+ realization: "nsrops.pow1_3"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_4:
+ realization: "nsrops.pow1_4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_5:
+ realization: "nsrops.pow1_5"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ sin:
+ realization: "numpy.sin"
+ alias: []
+ inverse: "asin"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ cos:
+ realization: "numpy.cos"
+ alias: []
+ inverse: "acos"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: false
+ monotonicity: 0
+ tan:
+ realization: "numpy.tan"
+ alias: []
+ inverse: "atan"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ asin:
+ realization: "numpy.arcsin"
+ alias: ["arcsin"]
+ inverse: "sin"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ acos:
+ realization: "numpy.arccos"
+ alias: ["arccos"]
+ inverse: "cos"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ atan:
+ realization: "numpy.arctan"
+ alias: ["arctan"]
+ inverse: "tan"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ exp:
+ realization: "numpy.exp"
+ alias: []
+ inverse: "log"
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ log:
+ realization: "numpy.log"
+ alias: ["ln"]
+ inverse: "exp"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 1
+
+variables: 3
\ No newline at end of file
diff --git a/configs/v10.0/nsr.yaml b/configs/v10.0/nsr.yaml
new file mode 100644
index 0000000..bc1939d
--- /dev/null
+++ b/configs/v10.0/nsr.yaml
@@ -0,0 +1,25 @@
+encoder_max_n_variables: 4 # includes the dependent variable
+size: 512
+
+pre_encoder_input_type: "ieee-754"
+pre_encoder_support_nan: False
+
+encoder: "SetTransformer2"
+encoder_kwargs:
+ hidden_size: 512
+ n_enc_isab: 5
+ n_dec_sab: 2
+ n_induce: 64
+ n_heads: 8
+ set_norm_ko: True
+ n_seeds: 64
+
+decoder_n_heads: 8
+decoder_ff_size: 512
+decoder_dropout: 0.1
+decoder_n_layers: 5
+
+learnable_positional_embeddings: False
+max_input_length: null
+
+expression_space: './expression_space.yaml'
\ No newline at end of file
diff --git a/configs/v10.0/skeleton_pool_train.yaml b/configs/v10.0/skeleton_pool_train.yaml
new file mode 100644
index 0000000..0683f6e
--- /dev/null
+++ b/configs/v10.0/skeleton_pool_train.yaml
@@ -0,0 +1,37 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/v7.0/skeleton_pool_val/",
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_proportional"
+ min_operators: 0
+ max_operators: 10
+ power: 1
+ max_length: 21
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v10.0/skeleton_pool_val.yaml b/configs/v10.0/skeleton_pool_val.yaml
new file mode 100644
index 0000000..29470da
--- /dev/null
+++ b/configs/v10.0/skeleton_pool_val.yaml
@@ -0,0 +1,36 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_proportional"
+ min_operators: 0
+ max_operators: 10
+ power: 1
+ max_length: 21
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v10.0/train.yaml b/configs/v10.0/train.yaml
new file mode 100644
index 0000000..6ecaa7a
--- /dev/null
+++ b/configs/v10.0/train.yaml
@@ -0,0 +1,28 @@
+model: ./nsr.yaml
+
+optimizer:
+ name: AdamW
+ kwargs:
+ lr: 1 # Will be multiplied by scheduler
+ weight_decay: 0.01
+ amsgrad: True
+
+lr_scheduler:
+ name: WarmupLinearAnnealing
+ kwargs:
+ min_lr: 0
+ max_lr: 1e-4
+ warmup_steps: 10_000
+ total_steps: 1_500_000
+
+batch_size: 128
+
+train_dataset: ./dataset_train.yaml
+val_dataset: "./dataset_val.yaml"
+val_batch_size: 128
+val_size: 100_000
+
+numeric_prediction_loss_weight: 0
+
+steps: 1_500_000
+device: cuda
\ No newline at end of file
diff --git a/configs/v11.0/dataset_train.yaml b/configs/v11.0/dataset_train.yaml
new file mode 100644
index 0000000..bb27322
--- /dev/null
+++ b/configs/v11.0/dataset_train.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_train.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v11.0/dataset_val.yaml b/configs/v11.0/dataset_val.yaml
new file mode 100644
index 0000000..c15ed03
--- /dev/null
+++ b/configs/v11.0/dataset_val.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_val.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v11.0/evaluation.yaml b/configs/v11.0/evaluation.yaml
new file mode 100644
index 0000000..98f6576
--- /dev/null
+++ b/configs/v11.0/evaluation.yaml
@@ -0,0 +1,14 @@
+n_support: 512
+beam_width: 32
+n_restarts: 8
+max_len: 32
+refiner_method: 'curve_fit_lm'
+numeric_head: False
+equivalence_pruning: True
+pointwise_close_criterion: 0.95
+pointwise_close_accuracy_rtol: 0.05
+pointwise_close_accuracy_atol: 0.001
+r2_close_criterion: 0.95
+refiner_p0_noise: 'uniform'
+refiner_p0_noise_kwargs: {'low': -5, 'high': 5}
+device: cuda
\ No newline at end of file
diff --git a/configs/v11.0/expression_space.yaml b/configs/v11.0/expression_space.yaml
new file mode 100644
index 0000000..8b17a96
--- /dev/null
+++ b/configs/v11.0/expression_space.yaml
@@ -0,0 +1,256 @@
+operators:
+ "+":
+ realization: "+"
+ alias: ["add", "plus"]
+ inverse: "-"
+ arity: 2
+ weight: 10
+ precedence: 1
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "-":
+ realization: "-"
+ alias: ["sub", "minus"]
+ inverse: "+"
+ arity: 2
+ weight: 5
+ precedence: 1
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ neg:
+ realization: "nsrops.neg"
+ alias: ["negative"]
+ inverse: "neg"
+ arity: 1
+ weight: 5
+ precedence: 2.5
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ "*":
+ realization: "*"
+ alias: ["mul", "times"]
+ inverse: "/"
+ arity: 2
+ weight: 10
+ precedence: 2
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "/":
+ realization: "/"
+ alias: ["div", "divide"]
+ inverse: "*"
+ arity: 2
+ weight: 5
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ abs:
+ realization: "abs"
+ alias: ["absolute"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ inv:
+ realization: "nsrops.inv"
+ alias: ["inverse"]
+ inverse: "inv"
+ arity: 1
+ weight: 4
+ precedence: 4
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ pow2:
+ realization: "nsrops.pow2"
+ alias: ["square"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ pow3:
+ realization: "nsrops.pow3"
+ alias: ["cube"]
+ inverse: "pow1_3"
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow4:
+ realization: "nsrops.pow4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 1
+ positive: true
+ monotonicity: 0
+ pow5:
+ realization: "nsrops.pow5"
+ alias: []
+ inverse: "pow1_5"
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_2:
+ realization: "nsrops.pow1_2"
+ alias: ["sqrt"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_3:
+ realization: "nsrops.pow1_3"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_4:
+ realization: "nsrops.pow1_4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_5:
+ realization: "nsrops.pow1_5"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ sin:
+ realization: "numpy.sin"
+ alias: []
+ inverse: "asin"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ cos:
+ realization: "numpy.cos"
+ alias: []
+ inverse: "acos"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: false
+ monotonicity: 0
+ tan:
+ realization: "numpy.tan"
+ alias: []
+ inverse: "atan"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ asin:
+ realization: "numpy.arcsin"
+ alias: ["arcsin"]
+ inverse: "sin"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ acos:
+ realization: "numpy.arccos"
+ alias: ["arccos"]
+ inverse: "cos"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ atan:
+ realization: "numpy.arctan"
+ alias: ["arctan"]
+ inverse: "tan"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ exp:
+ realization: "numpy.exp"
+ alias: []
+ inverse: "log"
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ log:
+ realization: "numpy.log"
+ alias: ["ln"]
+ inverse: "exp"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 1
+
+variables: 3
\ No newline at end of file
diff --git a/configs/v11.0/nsr.yaml b/configs/v11.0/nsr.yaml
new file mode 100644
index 0000000..bc1939d
--- /dev/null
+++ b/configs/v11.0/nsr.yaml
@@ -0,0 +1,25 @@
+encoder_max_n_variables: 4 # includes the dependent variable
+size: 512
+
+pre_encoder_input_type: "ieee-754"
+pre_encoder_support_nan: False
+
+encoder: "SetTransformer2"
+encoder_kwargs:
+ hidden_size: 512
+ n_enc_isab: 5
+ n_dec_sab: 2
+ n_induce: 64
+ n_heads: 8
+ set_norm_ko: True
+ n_seeds: 64
+
+decoder_n_heads: 8
+decoder_ff_size: 512
+decoder_dropout: 0.1
+decoder_n_layers: 5
+
+learnable_positional_embeddings: False
+max_input_length: null
+
+expression_space: './expression_space.yaml'
\ No newline at end of file
diff --git a/configs/v11.0/skeleton_pool_train.yaml b/configs/v11.0/skeleton_pool_train.yaml
new file mode 100644
index 0000000..0683f6e
--- /dev/null
+++ b/configs/v11.0/skeleton_pool_train.yaml
@@ -0,0 +1,37 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/v7.0/skeleton_pool_val/",
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_proportional"
+ min_operators: 0
+ max_operators: 10
+ power: 1
+ max_length: 21
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v11.0/skeleton_pool_val.yaml b/configs/v11.0/skeleton_pool_val.yaml
new file mode 100644
index 0000000..29470da
--- /dev/null
+++ b/configs/v11.0/skeleton_pool_val.yaml
@@ -0,0 +1,36 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_proportional"
+ min_operators: 0
+ max_operators: 10
+ power: 1
+ max_length: 21
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v11.0/train.yaml b/configs/v11.0/train.yaml
new file mode 100644
index 0000000..a418be9
--- /dev/null
+++ b/configs/v11.0/train.yaml
@@ -0,0 +1,21 @@
+model: ./nsr.yaml
+
+optimizer:
+ name: AdamWScheduleFree
+ kwargs:
+ lr: 0.0001
+ weight_decay: 0.01
+
+lr_scheduler: null
+
+batch_size: 128
+
+train_dataset: ./dataset_train.yaml
+val_dataset: "./dataset_val.yaml"
+val_batch_size: 128
+val_size: 100_000
+
+numeric_prediction_loss_weight: 0
+
+steps: 1_500_000
+device: cuda
\ No newline at end of file
diff --git a/configs/v7.11/dataset_train.yaml b/configs/v7.11/dataset_train.yaml
new file mode 100644
index 0000000..bb27322
--- /dev/null
+++ b/configs/v7.11/dataset_train.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_train.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v7.11/dataset_val.yaml b/configs/v7.11/dataset_val.yaml
new file mode 100644
index 0000000..c15ed03
--- /dev/null
+++ b/configs/v7.11/dataset_val.yaml
@@ -0,0 +1,2 @@
+skeleton_pool: './skeleton_pool_val.yaml'
+padding: 'zero'
\ No newline at end of file
diff --git a/configs/v7.11/evaluation.yaml b/configs/v7.11/evaluation.yaml
new file mode 100644
index 0000000..98f6576
--- /dev/null
+++ b/configs/v7.11/evaluation.yaml
@@ -0,0 +1,14 @@
+n_support: 512
+beam_width: 32
+n_restarts: 8
+max_len: 32
+refiner_method: 'curve_fit_lm'
+numeric_head: False
+equivalence_pruning: True
+pointwise_close_criterion: 0.95
+pointwise_close_accuracy_rtol: 0.05
+pointwise_close_accuracy_atol: 0.001
+r2_close_criterion: 0.95
+refiner_p0_noise: 'uniform'
+refiner_p0_noise_kwargs: {'low': -5, 'high': 5}
+device: cuda
\ No newline at end of file
diff --git a/configs/v7.11/expression_space.yaml b/configs/v7.11/expression_space.yaml
new file mode 100644
index 0000000..8b17a96
--- /dev/null
+++ b/configs/v7.11/expression_space.yaml
@@ -0,0 +1,256 @@
+operators:
+ "+":
+ realization: "+"
+ alias: ["add", "plus"]
+ inverse: "-"
+ arity: 2
+ weight: 10
+ precedence: 1
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "-":
+ realization: "-"
+ alias: ["sub", "minus"]
+ inverse: "+"
+ arity: 2
+ weight: 5
+ precedence: 1
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ neg:
+ realization: "nsrops.neg"
+ alias: ["negative"]
+ inverse: "neg"
+ arity: 1
+ weight: 5
+ precedence: 2.5
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ "*":
+ realization: "*"
+ alias: ["mul", "times"]
+ inverse: "/"
+ arity: 2
+ weight: 10
+ precedence: 2
+ commutative: true
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ "/":
+ realization: "/"
+ alias: ["div", "divide"]
+ inverse: "*"
+ arity: 2
+ weight: 5
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 0
+ abs:
+ realization: "abs"
+ alias: ["absolute"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ inv:
+ realization: "nsrops.inv"
+ alias: ["inverse"]
+ inverse: "inv"
+ arity: 1
+ weight: 4
+ precedence: 4
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: -1
+ pow2:
+ realization: "nsrops.pow2"
+ alias: ["square"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: true
+ monotonicity: 0
+ pow3:
+ realization: "nsrops.pow3"
+ alias: ["cube"]
+ inverse: "pow1_3"
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow4:
+ realization: "nsrops.pow4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 1
+ positive: true
+ monotonicity: 0
+ pow5:
+ realization: "nsrops.pow5"
+ alias: []
+ inverse: "pow1_5"
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_2:
+ realization: "nsrops.pow1_2"
+ alias: ["sqrt"]
+ inverse: null
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_3:
+ realization: "nsrops.pow1_3"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 2
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ pow1_4:
+ realization: "nsrops.pow1_4"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ pow1_5:
+ realization: "nsrops.pow1_5"
+ alias: []
+ inverse: null
+ arity: 1
+ weight: 1
+ precedence: 3
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ sin:
+ realization: "numpy.sin"
+ alias: []
+ inverse: "asin"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ cos:
+ realization: "numpy.cos"
+ alias: []
+ inverse: "acos"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 1 # symmetric
+ positive: false
+ monotonicity: 0
+ tan:
+ realization: "numpy.tan"
+ alias: []
+ inverse: "atan"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 0
+ asin:
+ realization: "numpy.arcsin"
+ alias: ["arcsin"]
+ inverse: "sin"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ acos:
+ realization: "numpy.arccos"
+ alias: ["arccos"]
+ inverse: "cos"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ atan:
+ realization: "numpy.arctan"
+ alias: ["arctan"]
+ inverse: "tan"
+ arity: 1
+ weight: 2
+ precedence: 2
+ commutative: false
+ symmetry: -1 # anti-symmetric
+ positive: false
+ monotonicity: 1
+ exp:
+ realization: "numpy.exp"
+ alias: []
+ inverse: "log"
+ arity: 1
+ weight: 4
+ precedence: 3
+ commutative: false
+ symmetry: 0
+ positive: true
+ monotonicity: 1
+ log:
+ realization: "numpy.log"
+ alias: ["ln"]
+ inverse: "exp"
+ arity: 1
+ weight: 4
+ precedence: 2
+ commutative: false
+ symmetry: 0
+ positive: false
+ monotonicity: 1
+
+variables: 3
\ No newline at end of file
diff --git a/configs/v7.11/nsr.yaml b/configs/v7.11/nsr.yaml
new file mode 100644
index 0000000..7ae31ee
--- /dev/null
+++ b/configs/v7.11/nsr.yaml
@@ -0,0 +1,25 @@
+encoder_max_n_variables: 4 # includes the dependent variable
+size: 512
+
+pre_encoder_input_type: "ieee-754"
+pre_encoder_support_nan: False
+
+encoder: "SetTransformer"
+encoder_kwargs:
+ hidden_size: 512
+ n_enc_isab: 5
+ n_dec_sab: 2
+ n_induce: 64
+ n_heads: 8
+ layer_norm: False
+ n_seeds: 64
+
+decoder_n_heads: 8
+decoder_ff_size: 512
+decoder_dropout: 0.1
+decoder_n_layers: 5
+
+learnable_positional_embeddings: False
+max_input_length: null
+
+expression_space: './expression_space.yaml'
\ No newline at end of file
diff --git a/configs/v7.11/skeleton_pool_train.yaml b/configs/v7.11/skeleton_pool_train.yaml
new file mode 100644
index 0000000..85031d5
--- /dev/null
+++ b/configs/v7.11/skeleton_pool_train.yaml
@@ -0,0 +1,38 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/v7.11/skeleton_pool_val/",
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_exponential"
+ min_operators: 0
+ max_operators: 15
+ power: 1
+ lambda: 1
+ max_length: 31
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v7.11/skeleton_pool_val.yaml b/configs/v7.11/skeleton_pool_val.yaml
new file mode 100644
index 0000000..250ee42
--- /dev/null
+++ b/configs/v7.11/skeleton_pool_val.yaml
@@ -0,0 +1,37 @@
+expression_space: './expression_space.yaml'
+holdout_pools: [
+ "{{ROOT}}/data/ansr-data/test_set/soose_nc/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/feynman/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/nguyen/skeleton_pool/",
+ "{{ROOT}}/data/ansr-data/test_set/pool_15/skeleton_pool/"
+]
+
+sample_strategy:
+ n_operator_distribution: "length_exponential"
+ min_operators: 0
+ max_operators: 15
+ power: 1
+ lambda: 1
+ max_length: 31
+ max_tries: 1
+ independent_dimensions: True
+
+allow_nan: False
+simplify: True
+
+literal_prior: 'uniform'
+literal_prior_kwargs:
+ low: -5
+ high: 5
+
+support_prior: "uniform_intervals"
+support_prior_kwargs:
+ low: -10
+ high: 10
+
+n_support_prior: "uniform"
+n_support_prior_kwargs:
+ low: 16
+ high: 512
+ min_value: 16
+ max_value: 512
\ No newline at end of file
diff --git a/configs/v7.11/train.yaml b/configs/v7.11/train.yaml
new file mode 100644
index 0000000..6ecaa7a
--- /dev/null
+++ b/configs/v7.11/train.yaml
@@ -0,0 +1,28 @@
+model: ./nsr.yaml
+
+optimizer:
+ name: AdamW
+ kwargs:
+ lr: 1 # Will be multiplied by scheduler
+ weight_decay: 0.01
+ amsgrad: True
+
+lr_scheduler:
+ name: WarmupLinearAnnealing
+ kwargs:
+ min_lr: 0
+ max_lr: 1e-4
+ warmup_steps: 10_000
+ total_steps: 1_500_000
+
+batch_size: 128
+
+train_dataset: ./dataset_train.yaml
+val_dataset: "./dataset_val.yaml"
+val_batch_size: 128
+val_size: 100_000
+
+numeric_prediction_loss_weight: 0
+
+steps: 1_500_000
+device: cuda
\ No newline at end of file
diff --git a/configs/v9.0/dataset_train.yaml b/configs/v9.0/dataset_train.yaml
index bb27322..02faf42 100644
--- a/configs/v9.0/dataset_train.yaml
+++ b/configs/v9.0/dataset_train.yaml
@@ -1,2 +1,6 @@
skeleton_pool: './skeleton_pool_train.yaml'
-padding: 'zero'
\ No newline at end of file
+padding: 'zero'
+preprocessor:
+ expression_space: './expression_space.yaml'
+ format_probs:
+ complexity: 0.5
\ No newline at end of file
diff --git a/configs/v9.0/dataset_val.yaml b/configs/v9.0/dataset_val.yaml
index c15ed03..ddc9d07 100644
--- a/configs/v9.0/dataset_val.yaml
+++ b/configs/v9.0/dataset_val.yaml
@@ -1,2 +1,6 @@
skeleton_pool: './skeleton_pool_val.yaml'
-padding: 'zero'
\ No newline at end of file
+padding: 'zero'
+preprocessor:
+ expression_space: './expression_space.yaml'
+ format_probs:
+ complexity: 0.5
\ No newline at end of file
diff --git a/configs/v9.0/nsr.yaml b/configs/v9.0/nsr.yaml
index 7ae31ee..d7150f3 100644
--- a/configs/v9.0/nsr.yaml
+++ b/configs/v9.0/nsr.yaml
@@ -21,5 +21,6 @@ decoder_n_layers: 5
learnable_positional_embeddings: False
max_input_length: null
+support_numeric_tokens: True
expression_space: './expression_space.yaml'
\ No newline at end of file
diff --git a/experimental/control_formatter.ipynb b/experimental/control_formatter.ipynb
index bb52a7b..8927fb8 100644
--- a/experimental/control_formatter.ipynb
+++ b/experimental/control_formatter.ipynb
@@ -6,8 +6,7 @@
"metadata": {},
"outputs": [],
"source": [
- "from flash_ansr.control import ControlFormatter\n",
- "from flash_ansr import FlashANSRDataset, get_path, Tokenizer"
+ "from flash_ansr import FlashANSRDataset, get_path"
]
},
{
@@ -19,78 +18,150 @@
"name": "stderr",
"output_type": "stream",
"text": [
- "Compiling Skeletons: 100%|██████████| 200/200 [00:00<00:00, 45729.44it/s]\n",
- "Compiling Skeletons: 100%|██████████| 43/43 [00:00<00:00, 41327.93it/s]\n",
- "Compiling Skeletons: 100%|██████████| 10/10 [00:00<00:00, 26903.81it/s]\n",
- "Compiling Skeletons: 100%|██████████| 4999/4999 [00:00<00:00, 25354.31it/s]\n"
+ "Compiling Skeletons: 100%|██████████| 200/200 [00:00<00:00, 39919.14it/s]\n",
+ "Compiling Skeletons: 100%|██████████| 43/43 [00:00<00:00, 38202.73it/s]\n",
+ "Compiling Skeletons: 100%|██████████| 10/10 [00:00<00:00, 23458.08it/s]\n",
+ "Compiling Skeletons: 100%|██████████| 4999/4999 [00:00<00:00, 26677.01it/s]\n"
]
}
],
"source": [
- "dataset = FlashANSRDataset.from_config(get_path('configs', 'v7.0', 'dataset_val.yaml'))"
+ "dataset = FlashANSRDataset.from_config(get_path('configs', 'v9.0', 'dataset_val.yaml'))"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
- "outputs": [],
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "dict_keys(['n_rejected', 'skeletons', 'skeleton_hashes', 'expressions', 'constants', 'input_ids', 'x_tensors', 'y_tensors'])\n",
+ "[[(1, 7)], []]\n",
+ "tensor([[ 7, 6, 1, 11, 6, 12, 32, 33, 2, 0, 0, 0, 0],\n",
+ " [ 1, 9, 11, 6, 8, 31, 32, 8, 6, 21, 19, 32, 2]])\n",
+ "['FlashANSR(beam_width=32,\n", + " expression_space=<flash_ansr.expressions.expression_space.ExpressionSpace object at 0x7fe04f2140d0>,\n", + " flash_ansr_transformer=FlashANSRTransformer(\n", + " (pre_encoder): PreEncoder()\n", + " (encoder): SetTransformer(\n", + " (enc): Sequential(\n", + " (0): ISAB(\n", + " (mab0): MAB(\n", + " (W_q): Linear(in_features=512, out_features=512, bias=True)\n", + " (W_k): Linear(in_features=64, out_features=...\n", + " (0): Linear(in_features=512, out_features=512, bias=True)\n", + " (1): GELU(approximate='none')\n", + " (2): Dropout(p=0.1, inplace=False)\n", + " (3): Linear(in_features=512, out_features=33, bias=True)\n", + " )\n", + " (num_out): Sequential(\n", + " (0): Linear(in_features=512, out_features=512, bias=True)\n", + " (1): GELU(approximate='none')\n", + " (2): Dropout(p=0.1, inplace=False)\n", + " (3): Linear(in_features=512, out_features=1, bias=True)\n", + " )\n", + "),\n", + " n_restarts=8, verbose=True)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
FlashANSR(beam_width=32,\n", + " expression_space=<flash_ansr.expressions.expression_space.ExpressionSpace object at 0x7fe04f2140d0>,\n", + " flash_ansr_transformer=FlashANSRTransformer(\n", + " (pre_encoder): PreEncoder()\n", + " (encoder): SetTransformer(\n", + " (enc): Sequential(\n", + " (0): ISAB(\n", + " (mab0): MAB(\n", + " (W_q): Linear(in_features=512, out_features=512, bias=True)\n", + " (W_k): Linear(in_features=64, out_features=...\n", + " (0): Linear(in_features=512, out_features=512, bias=True)\n", + " (1): GELU(approximate='none')\n", + " (2): Dropout(p=0.1, inplace=False)\n", + " (3): Linear(in_features=512, out_features=33, bias=True)\n", + " )\n", + " (num_out): Sequential(\n", + " (0): Linear(in_features=512, out_features=512, bias=True)\n", + " (1): GELU(approximate='none')\n", + " (2): Dropout(p=0.1, inplace=False)\n", + " (3): Linear(in_features=512, out_features=1, bias=True)\n", + " )\n", + "),\n", + " n_restarts=8, verbose=True)
| \n", + " | log_prob | \n", + "fvu | \n", + "score | \n", + "expression | \n", + "complexity | \n", + "target_complexity | \n", + "numeric_prediction | \n", + "beam | \n", + "function | \n", + "refiner | \n", + "beam_id | \n", + "fit_constants | \n", + "fit_covariances | \n", + "fit_loss | \n", + "
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | \n", + "-1.060831 | \n", + "1.152357e-15 | \n", + "0.0007 | \n", + "[+, x1, +, x2, *, <num>, x3] | \n", + "7 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 31, 10, 6, 32, 2] | \n", + "<function <lambda> at 0x7fe04ea5b600> | \n", + "Refiner(expression=['+', 'x1', '+', 'x2', '*',... | \n", + "0 | \n", + "[2.9999999960488646] | \n", + "[[1.3347598938741106e-16]] | \n", + "4.053961e-13 | \n", + "
| 1 | \n", + "-1.060831 | \n", + "1.152357e-15 | \n", + "0.0007 | \n", + "[+, x1, +, x2, *, <num>, x3] | \n", + "7 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 31, 10, 6, 32, 2] | \n", + "<function <lambda> at 0x7fe04ea5b600> | \n", + "Refiner(expression=['+', 'x1', '+', 'x2', '*',... | \n", + "0 | \n", + "[2.9999999960488646] | \n", + "[[1.33475989113999e-16]] | \n", + "4.053961e-13 | \n", + "
| 2 | \n", + "-1.060831 | \n", + "1.152357e-15 | \n", + "0.0007 | \n", + "[+, x1, +, x2, *, <num>, x3] | \n", + "7 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 31, 10, 6, 32, 2] | \n", + "<function <lambda> at 0x7fe04ea5b600> | \n", + "Refiner(expression=['+', 'x1', '+', 'x2', '*',... | \n", + "0 | \n", + "[2.9999999960488646] | \n", + "[[1.3347598871570362e-16]] | \n", + "4.053961e-13 | \n", + "
| 3 | \n", + "-1.060831 | \n", + "1.152357e-15 | \n", + "0.0007 | \n", + "[+, x1, +, x2, *, <num>, x3] | \n", + "7 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 31, 10, 6, 32, 2] | \n", + "<function <lambda> at 0x7fe04ea5b600> | \n", + "Refiner(expression=['+', 'x1', '+', 'x2', '*',... | \n", + "0 | \n", + "[2.9999999960488646] | \n", + "[[1.3347598928734102e-16]] | \n", + "4.053961e-13 | \n", + "
| 4 | \n", + "-1.060831 | \n", + "1.152357e-15 | \n", + "0.0007 | \n", + "[+, x1, +, x2, *, <num>, x3] | \n", + "7 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 31, 10, 6, 32, 2] | \n", + "<function <lambda> at 0x7fe04ea5b600> | \n", + "Refiner(expression=['+', 'x1', '+', 'x2', '*',... | \n", + "0 | \n", + "[2.9999999960488646] | \n", + "[[1.3347599027477214e-16]] | \n", + "4.053961e-13 | \n", + "
| ... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "... | \n", + "
| 237 | \n", + "-6.061850 | \n", + "3.152573e-15 | \n", + "0.0013 | \n", + "[+, <num>, +, <num>, +, x1, +, x2, +, x3, +, x... | \n", + "13 | \n", + "None | \n", + "None | \n", + "[1, 7, 6, 7, 6, 7, 30, 7, 31, 7, 32, 7, 32, 32... | \n", + "<function <lambda> at 0x7fe04ea58c20> | \n", + "Refiner(expression=['+', '<num>', '+', '<num>'... | \n", + "30 | \n", + "[-0.47078499131995105, 0.47078492063308375] | \n", + "[[0.07722945023551932, -0.07722946672644812], ... | \n", + "1.109068e-12 | \n", + "
| 238 | \n", + "-6.061850 | \n", + "3.152573e-15 | \n", + "0.0013 | \n", + "[+, <num>, +, <num>, +, x1, +, x2, +, x3, +, x... | \n", + "13 | \n", + "None | \n", + "None | \n", + "[1, 7, 6, 7, 6, 7, 30, 7, 31, 7, 32, 7, 32, 32... | \n", + "<function <lambda> at 0x7fe04ea58c20> | \n", + "Refiner(expression=['+', '<num>', '+', '<num>'... | \n", + "30 | \n", + "[0.5407137010650169, -0.5407137717518034] | \n", + "[[1.468625096212761, -1.4686252416604124], [-1... | \n", + "1.109068e-12 | \n", + "
| 239 | \n", + "-6.061850 | \n", + "3.152573e-15 | \n", + "0.0013 | \n", + "[+, <num>, +, <num>, +, x1, +, x2, +, x3, +, x... | \n", + "13 | \n", + "None | \n", + "None | \n", + "[1, 7, 6, 7, 6, 7, 30, 7, 31, 7, 32, 7, 32, 32... | \n", + "<function <lambda> at 0x7fe04ea58c20> | \n", + "Refiner(expression=['+', '<num>', '+', '<num>'... | \n", + "30 | \n", + "[0.3074403724255982, -0.30744044311224167] | \n", + "[[0.01270703826740337, -0.012707037074853388],... | \n", + "1.109068e-12 | \n", + "
| 240 | \n", + "-6.061850 | \n", + "3.152573e-15 | \n", + "0.0013 | \n", + "[+, <num>, +, <num>, +, x1, +, x2, +, x3, +, x... | \n", + "13 | \n", + "None | \n", + "None | \n", + "[1, 7, 6, 7, 6, 7, 30, 7, 31, 7, 32, 7, 32, 32... | \n", + "<function <lambda> at 0x7fe04ea58c20> | \n", + "Refiner(expression=['+', '<num>', '+', '<num>'... | \n", + "30 | \n", + "[0.31953753768780313, -0.3195376083731668] | \n", + "[[0.060332209908029115, -0.06033219583661418],... | \n", + "1.109068e-12 | \n", + "
| 241 | \n", + "-6.257017 | \n", + "6.773130e-15 | \n", + "0.0013 | \n", + "[+, x1, +, x1, +, x2, +, x3, +, x3, -, x3, x1] | \n", + "13 | \n", + "None | \n", + "None | \n", + "[1, 7, 30, 7, 30, 7, 31, 7, 32, 7, 32, 8, 32, ... | \n", + "<function <lambda> at 0x7fe04ea5b2e0> | \n", + "Refiner(expression=['+', 'x1', '+', 'x1', '+',... | \n", + "31 | \n", + "[] | \n", + "[] | \n", + "2.382770e-12 | \n", + "
242 rows × 14 columns
\n", + "